Knn learns composition of two Bach's original works and then stored into a Cloud system for later retrieval. A simple tune was fed to the Learned cloud function and improv-ed a new tune!Wolfram Mathematica code:https://www.wolframcloud.com/obj/ccn2/Published/learned_function_air_adagio.nb
New Algorithmic Improv:Wolfram Mathematica code:https://www.wolframcloud.com/obj/ccn2/Published/knn_partition_similarity_air.nb
New Algorithmic Improv:Wolfram Mathematica code:https://www.wolframcloud.com/obj/ccn2/Published/knn_partition_similarity_air.nb
New Algorithmic Improv:Wolfram Mathematica code:https://www.wolframcloud.com/obj/ccn2/Published/knn_partition_similarity_air.nb
New Algorithmic Improv:Wolfram Mathematica code:https://www.wolframcloud.com/obj/ccn2/Published/knn_partition_similarity_air.nb
About 10 years back, I needed repeatedly to listen to Art of Fugue (AOF) and problem became patent that Classical Chamber performances were acoustically too mangled to listen to different passages/voices independently.
Listened to organ performances, though much better left much more to desire.
So I took a small piece and rearranged the instruments to be able to listen more attentively, in order to learn the art of Fugue composition.
And as I always I had to throw in kitchen sink full of synthesizers and rock drums and basses and yes Indian Tabla!
Haven't you heard! Mirror fallen in love with his own reflection!
Production Notes:
I call this style Pareh-Pareh, Farsi for (torn) piece by piece, the idea is to stitch together pieces different in styles and arrangements and yet in composer's butchered heart they are manifestation of a singleton love.
Hasten attempts to stitch pieces by the thread and needle of sorrows and regrets, yet all the composer endeavours ... some noises.
Cover Photo by Fares Hamouche
https://www.vfrhub.com/article/re-experiencing-symptom-self-monitoring-as-an-early-step-in-the-treatment-of-posttraumatic-stress-disorder/fares-hamouche-dv7rtvn_c6q-unsplash/#
Derivative Image Processing by Dara.
No flames too hot for a true lover who braves infernos, in search of the Beloved, and yet once his ashes arrive, what song! can ashes serenade the Beloved?
And if you do find a song within, you have not reached the Beloved.
A Darvish's gift for the good friend Joseph Umo, The Sorrower.
A testament to the sorrows of all Persians:
You made night, I made candle You made dust, I made goblet Deserts and mountains and valleys you made Streets and meadows and orchards I made I! who from the stone makes shiny mirror I! who from the toxin makes sweet drink
The original Farsi poem and the actual are song are not the works of Dara. The Translation and arrangement are.
https://soundcloud.com/mahsaesfand/ghazal-shakeri-song-of-red-dervish
شب آفریدی، شمع آفریدم خاک آفریدی، جام آفریدم بیابان و کوهسار و راغ آفریدی خیابان و گلزار و باغ آفریدم آنم که از سنگ آیینه سازم آنم که از زهر نوشینه سازم
Bach's Air is grafted into Walkingears' 2 50 9 waltz.
The notes are from Air, the closest groups of 40 notes yet the timing is still the original Walkingears.
Distance function CanberraDistance, 40NN Similarity algorithm
My Revolutionary hero https://www.youtube.com/watch?v=Nysdq_plUTg
Chopin's Revolutionary (slowed down) was partitioned into groups of 19 midi notes and 15NN learned the partition and improvised by outputting the second most similar 19 midi partition from the same Revolutionary composition.
Bach's Air was partitioned into groups of 40 midi notes and 2NN learned the partition and improvised by outputting the second most similar 40 partition from the same Air composition.
Surprisingly much of the original sensual mood was preserved and yet the composition is discernibly a new one while the listener wonders where the Bach's Air come into it all...
The new KNN algorithm (k=8) learns a composition in groups or vectors of 4 midi events. It then takes these groups of 5 and re-arranges them i.e. combinatorics while preserving the original timing. As a result very rich and enjoyable tunes are created.
FIXME: fix the drums and kicks in particular.
Collaborative work with WalkingEars. WalkingEars wrote a beautiful waltz and I wrote the AI Improv algorithm which learned his composition and composed a new one as you can hear.
The new algorithm learns your composition in groups or vectors of 5 midi events. It then takes these groups of 5 and re-arranges them i.e. combinatorics while preserving the original timing. As a result very interesting and enjoyable tunes are created.
K-nn n=43 CanberraDistance learned the Fugue in D minor and then improvised.
Had Bach and Rumi met...
k-nn learned Toccata D minor and then applied to the same midi and improvised a new version.
n=43 BrayCurtisDistance, 43rd nearest note was computed with BrayCrurtis Distance Function.
Cover: Neural Net brush style repainted Bach's portrait.
https://www.wassilykandinsky.net/work-370.php
Toccata D minor k-nn learned with Cosine Distance Function. The improve computed by apply the learned algorithm to the original composition.
Bach portrait was re-styled by Van Gogh's cherry blossom brush.
Midi http://files.untiredwithloving.org/knn_toccataD_cosine_midi.zip
midi transcriptions of Air on G String and BWV 974 were interpolated via Sigmoidal interpolant.
Improv starts with Air and then fuses into BWV 974 until it gets some traction (or sound like the original) and then fuses back to Air.
midi transcriptions of Air on G String and BWV 974 were interpolated via Sigmoidal interpolant.
Midi timing for BWV 1068 preserved
The frequency/note obtained by linear combination of k-nn (nearest in the BWV 1068, k=5 or 10) + original frequency
The left hand was shifted to earlier time and thus right hand later
Make the k-nn similar enough so some of the tunes of the original composition could be heard and yet it is entirely new.
I like to experiment with these improvs but take out smaller pieces and make different parts asynchronous to get better creative sounds.
After making this, I was listening to "Play Bach The Jacgues Loussier Trio" and interestingly found some similar sounding improvs in his.
"Lover is the redolent incense burning on fire" (Ibn Manzour Afriqi)
BWV 974(original) https://www.youtube.com/watch?v=2x-OHljZzHQ
BWV 974 midi-timings stripped and new note frequencies obtained by computing the Nearest note in Chopin Revolutionary+Tico Tico+ BWV 974 (linear combination), normalized by 1/3.
New AI improve to get better timing to avoid the super fast notes previously difficult to listen to.
K-nn algorithm computed the nearest notes in BWV 988 (Aria) to notes in Chopin's Revolutionary and then mixed them again with somewhat altered timing of Revolutionary.
The repeated counterpoint is the real composition, not the melodies on the piano. Some plangent vynil synth sounded the counterpoint.
The idea is to piece together broken shards found NEAR old classics, but yet so different.
Piano ornamentals were added by hand and panned to left channel. Their purpose was to rendered a one way discourse with a beloved that is not patently replying! Therefore, there is a sense of vying for response that never arrives.
"Some evening when moon was my talking The beauty of your face, celestial came to mind"
All that you see, all that you hear, all that you think are but infinite reflections of Singleton Beauty.
"Deep in darkness of night was my setting Curls of your hair, sable came to mind"
When you said to her “I love you”, who did you actually love?
"Rolling tears betwixt the dust and blood was my seeing Martyrs of your enchanted eyes, copious came to mind"
Poetry by Kamal Khujandi https://en.wikipedia.org/wiki/Kamal_Khujandi
And while in throes of happiness and fortune, or midst the gore and mayhem, what was it that you stared at?
Whilst bemoaning her loss, which beloved did you really yearn towards?
Production Notes:
Toccata: When you said to her “I love you”, which beloved did you actually speak to? Used the vocals to emulate a lovers quarrel, per style of Toccata. Originally was planning to use the Cathedral organs, but then it came to me Toccata is really for human voice and rock lead guitars!
The base melody is from the Persian phrase:
Doustat Daaram (I love you)
Most of the counterpoints were mirror reflections of a few notes from the melody.
Drums added inspired by the drummer of this Toccata in D minor arrangement https://youtu.be/QgbgUrp1a70
Where there is Beloved, when there is Beloved, how there is Beloved, there is no me nor I nor you nor us and no them!
Improv on Yamaha P-125.
Instrumentation and arrangement by LOGIC PRO.
Research
Listen to this Nay (Reed-flute) how it complains How it tells the tales of (long lasting) separations
Since I have been chopped off the marsh-reed Men and women moan listening to my screams
http://www.untiredwithloving.org/masnavi_1_1.html
BWV847 https://www.youtube.com/watch?v=3ceiQadblRQ
K-nn finds the 100th nearest note (f2) for each note (f1) in BWV847, then improvises according to interpolation:
new note = (1-x)f1 + xf2 start time=(1-y)t1start + yt1end t1 for f1 end time=(1-y)t2star+ yt2end t2 for f2 x=0.75 y=0.001
x is large since I like it to all together a different sound and yet carry some bits from Bach's. y is tiny just a dash of time variation.
Instrumentation Logic PRO X Alchemy, Ve-Va-Vo phones used to chant the new improvisation.
What is Melodic Forage? There are those who compose because their spirits are famished and languid under the burden of their own darkenesses; their souls forage for acoustic light like unto Honey bees foraging for pollen and yet aflight they pollinate other souls with de novo thoughts
Image sonification source: https://www.nikonsmallworld.com/galleries/entry/2013-photomicrography-competition/1
Production Code: http://files.untiredwithloving.org/Chaetoceros_debilis.pdf
https://youtu.be/KTcJHfDq7gE
Hydrangea single bloom scanned vertically from left to write, strummed the vertical scans, sonified and midified.
AI trained with Hydrangea midi (K-nn) then improvised Gavotte in Partita 6 (BMV 830) into a new composition though parts are still familiar. Weighted average of the Gavotte and the AI improv computed and midified.
Logic PRO for instrumentation.
K-NN learned BWV773 and then entirely computed two distinct voice polyphonic composition.
Voice1: K-NN computing the at most 10 closest neighbours (to each midi event)of radius between 10 to 30 (10<= r <=30)
Voice1: K-NN computing the at most 5 closest neighbours of radius between 8 to 30 (8<= r <=30)
References: https://www.youtube.com/watch?v=qDroecI3270 https://www.youtube.com/watch?v=UqYde-LULfs
Sufi chant of Basmala turned into midi by Melodyne. Instrumentation by Logic Pro. Two voices, both from the same source.
Chanter El-Kandari.
Reference: https://soundcloud.com/dara-o-shayda/darwishaan-sob-at-new-crater-impact
Quf Chant by Kandari.
EXS24 sampler built from the voice chant, and the reversed voice chant.
Rotifer Edge Detect filter
See Also https://soundcloud.com/dara-o-shayda/darwishaans-rotifer-fugue-dance-in-chamber-of-non-being
Rotifer a fully functional being has about 250 neurons!
This music is not just about emoting but mentation as well.
https://www.youtube.com/watch?v=wDyKhhqP0Xc
https://en.wikipedia.org/wiki/Rotifer
Production notes:
Image of the Rotifer was transformed using the Skeleton Transform: http://reference.wolfram.com/language/ref/SkeletonTransform.html
Scanned vertically from left to right and sonified in Mathematica 11.2. Logic Pro instrumentation.
http://files.untiredwithloving.org/Rotifer.pdf
Skeletal Transform: https://reference.wolfram.com/language/ref/SkeletonTransform.html
I use this transformation to compute melodies that are more recital like for Piano instrumentations.
See for details: https://soundcloud.com/dara-o-shayda/darwishaan-sob-at-new-crater-impact
Basically the image for was transformed by Skeletal Transform, sonified vertically from left to right. Melodyne used to midify and it was pitch and time corrected and snapped to scale B Major. Instrumentations by Logic PRO.
Source:https://www.nasa.gov/image-feature/jpl/pia21459/a-new-crater-on-a-dusty-slope
Image, the crater, scanned vertically from from left to right. Sonified in Mathematica using AAVAA. Melodyne converted audio to midi. Instrumentation with Logic Pro.
Production Code: http://files.untiredwithloving.org/newcrater.pdf
Source: http://d3lp4xedbqa8a5.cloudfront.net/s3/digital-cougar-assets/AusGeo/2016/11/03/64021/5-Siwanowicz_1-(Custom).jpg
Front foot (tarsus) of a male diving beetle, Confocal, 100x Credit: Dr. Igor Siwanowicz (Howard Hughes Medical Institute, USA)
http://www.pnas.org/content/97/12/6568.full
Image processed by Edge Detect and scanner horizontally from left to right and sonified pixel by pixel; turned into midi by Melodyne (poly). Logic Pro human voice synthesizers, all trumpet horn sounds are human voice in low frequencies.
Image filtered by the image pattern extraction:
https://reference.wolfram.com/language/ref/MorphologicalPerimeter.html
The resultant was scanned vertically and swept from left to right and sonified pixel by pixel. Then midified by the Melodyne and post processing done in Logic Pro X, alchemy with custom arpeggio.
Production code: http://files.lossofgenerality.com/dew.pdf
Image: http://www.krebsmicro.com/forumpix/7278-7302small.jpg
Research: https://en.wikipedia.org/wiki/Diatom
Production Notes:
All post production done in Logic Pro X, the chorals are by Alchemy. http://files.lossofgenerality.com/diatom.pdf
See how the melodies extracted from the image:
https://soundcloud.com/dara-o-shayda/stylophora
Source: http://mgl.scripps.edu/people/goodsell/illustration/public
Cytoplasm sketch vertically scanned from left to right and sonified by AAvAA a Mathematica application for image sonification. Production in Logic Pro.
Source: http://www.nikonsmallworld.com/galleries/entry/2015-photomicrography-competition/34
Image of Distinction 2015 Photomicrography Competition Dr. Kesara Margrét Anamthawat-Jonsson, Andrey N. Gagunashvili & Ólafur S. Andrésson University of Iceland Reykjavik, Iceland
Subject Matter:
Nostoc, a blue-green alga (cyanobacteria) showing red autofluorescence of the chlorophylls (400x)
Production Notes:
Image vertically scanned from left to right, sonified by AAVAA a Mathematica image sonifier. Midified by Melodyne (POLY). Counterpoint in Logic PRO.
Production Code:
http://files.lossofgenerality.com/nostoc.pdf
Image EdgeDetected, then scanned vertically from left to right, the sonified, and midified by Melodyone. Instrumentation and arrangement in Logic PRO.
Production Code: http://files.lossofgenerality.com/otolih.pdf
See: https://humanphysiology2011.wikispaces.com/file/view/the_otolith_organ.jpg/218917154/487x516/the_otolith_organ.jpg
https://classconnection.s3.amazonaws.com/184/flashcards/1904184/jpg/picture61352569073487.jpg
http://www.dizziness-and-balance.com/disorders/bppv/otoliths.html
Graph Theory algorithm called Shortest Tour (variation of Traveling Salesman)
en.wikipedia.org/wiki/Travelling_salesman_problem
was applied to Black&White photo of Monalisa and then a simple synthesizer was applied to each step of the tour (travel). 1000 steps chosen from the tour and sonified accordingly.
Two melodies, 3 voices, in counterpoint, one Shortest Tour for white part of the photo, the other Shorter Tour for the black part of the photo.
Instrumentation in Logic PRO, Piano Motion for two voices (right, left), and Celtic Harp the third.
Remark: No special effects used for the low frequency spacy sounds, they are the original sampled instruments.
Production code: http://files.lossofgenerality.com/monaBW.pdf
See Also:
https://soundcloud.com/dara-o-shayda/notch-lamenting-delta-positive-skeleton
Arpeggiator derived by the Notch Active Skeleton Transform.
Source: www.esa.int/spaceinimages/Images/2015/04/Saturn_s_sponge-like_moon_Hyperion
Image transformed by Morphological Perimeter algoritm:
https://reference.wolfram.com/language/ref/MorphologicalPerimeter.html
scanned vertically from left to right, midified by Melodyne (Mono). The mom-midification was then echoed higher (a copy of the midi moved to the right of the keyboard 1 note) with a slight time-shift (see the production code).
Counter Point arrangement in Logic Pro. Each counter point was arpeggiated, three speeds used with two different patterns. First pattern opens the composition with lowest speed, then the same pattern repeats (counterpoints) each time with higher speeds. The beginning two counterpoints again initiated in the middle of fugue.
Production code and notes:
http://files.lossofgenerality.com/Hyperion.pdf
Source: www.esa.int/spaceinimages/Images/2015/04/Saturn_s_sponge-like_moon_Hyperion
Image transformed by Morphological Perimeter algoritm:
https://reference.wolfram.com/language/ref/MorphologicalPerimeter.html
scanned vertically from left to right, midified by Melodyne (Poly). Counter Point arrangement in Logic Pro. Each counter point was arpeggiated, three speeds used with two different patterns. First pattern opens the composition with lowest speed, then the same pattern repeats (counterpoints) each time with higher speeds. The beginning two counterpoints again initiated in the middle of fugue.
Production code and notes:
http://files.lossofgenerality.com/Hyperion.pdf
Source: www.esa.int/spaceinimages/Images/2015/04/Saturn_s_sponge-like_moon_Hyperion
Image transformed by Morphological Perimeter algoritm:
https://reference.wolfram.com/language/ref/MorphologicalPerimeter.html
scanned vertically from left to right, midified by Melodyne (Poly). Counter Point arrangement in Logic Pro. Each counter point was arpeggiated, two speeds used with two different patterns.
Production code and notes:
http://files.lossofgenerality.com/Hyperion.pdf
One of the Dubliners' pianists had accused me of composing dark tunes, so this was a response :)
Source:http://www.riken.jp/en/research/rikenresearch/highlights/7994/
Notch-Active(green) separated from Delta-Positive (red), sonified scanned vertical from left to right, each assigned to left and right channels accordingly. Skeleton Transform applied to compute the medial axis and metric distance to background:
http://reference.wolfram.com/language/ref/SkeletonTransform.html
Midified by Melodyne (POLY). Improvised by muting some notes, volume change for others. Arrangements in Logic PRO, Ghost Piano for Red (Delta-Positive).
Source:http://www.riken.jp/en/research/rikenresearch/highlights/7994/
Notch-Active(green) separated from Delta-Positive (red), sonified scanned vertical from left to right, each assigned to left and right channels accordingly. Midified by Melodyne (POLY). Improvised by muting some notes. Arrangements in Logic PRO, Chorus for Red (Delta-Positive), Transistor Radio Piano for Green (Notch Active).
I usually choose Transistor Radio Piano to render infinite distance in time or space.
Source:http://www.riken.jp/en/research/rikenresearch/highlights/7994/
Notch-Active(green) separated from Delta-Positive (red), sonified scanned vertical from left to right, each assigned to left and right channels accordingly. Midified by Melodyne (POLY). Improvised by muting some notes. Arrangements in Logic PRO.
Code and images:
http://files.lossofgenerality.com/notch.pdf